Abstract

This paper describes an optical implementation of a winner-take-all (WTA) function using a VLSI/liquid crystal spatial light modulator. The WTA function evaluates an array of continuous valued inputs and produces a binary output vector with components that are true for the largest of the inputs (the winner) and false for the others. The WTA function is used in neural network models such as competitive learning and for associative memories and correlation plane detectors. Optical implementations of these systems require a WTA with optical inputs and outputs. VLSI/liquid crystal devices provide photodetectors for optical inputs, liquid crystal modulators for optical outputs, and electronic circuitry to implement the globally nonlinear WTA function.

© 1991 Optical Society of America